Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170665
Choon-seng Yee, K. Lim
The parallel manipulator's unique structure presents an interesting problem in its forward kinematics solution, which involves the solving of a series of simultaneous nonlinear equations. The ability of a neural network to recognize the relationship between the input values and the output values of a system without fully understanding the system was fully exploited in this case. With the simple inverse kinematics solution of the manipulator, a neural network was trained to solve the forward kinematics of the parallel manipulator quite accurately. By adjusting the offset of the result obtained, the neural network is able to achieve an accuracy of 0.1 mm and 0.5 degrees for the six output values.<>
{"title":"Neural network for the forward kinematics problem in parallel manipulator","authors":"Choon-seng Yee, K. Lim","doi":"10.1109/IJCNN.1991.170665","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170665","url":null,"abstract":"The parallel manipulator's unique structure presents an interesting problem in its forward kinematics solution, which involves the solving of a series of simultaneous nonlinear equations. The ability of a neural network to recognize the relationship between the input values and the output values of a system without fully understanding the system was fully exploited in this case. With the simple inverse kinematics solution of the manipulator, a neural network was trained to solve the forward kinematics of the parallel manipulator quite accurately. By adjusting the offset of the result obtained, the neural network is able to achieve an accuracy of 0.1 mm and 0.5 degrees for the six output values.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114691414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170480
Butong Zhang, G. Veenker
The authors introduce an active learning paradigm for neural networks. In contrast to the passive paradigm, the learning in the active paradigm is initiated by the machine learner instead of its environment or teacher. The authors present a learning algorithm that uses a genetic algorithm for creating novel examples to teach multilayer feedforward networks. The creative learning networks, based on their own knowledge, discover new examples, criticize and select useful ones, train themselves, and thereby extend their existing knowledge. Experiments on function extrapolation show that the self-teaching neural networks not only reduce the teaching efforts of the human, but the genetically created examples also contribute robustly to the improvement of generalization performance and the interpretation of the connectionist knowledge.<>
{"title":"Neural networks that teach themselves through genetic discovery of novel examples","authors":"Butong Zhang, G. Veenker","doi":"10.1109/IJCNN.1991.170480","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170480","url":null,"abstract":"The authors introduce an active learning paradigm for neural networks. In contrast to the passive paradigm, the learning in the active paradigm is initiated by the machine learner instead of its environment or teacher. The authors present a learning algorithm that uses a genetic algorithm for creating novel examples to teach multilayer feedforward networks. The creative learning networks, based on their own knowledge, discover new examples, criticize and select useful ones, train themselves, and thereby extend their existing knowledge. Experiments on function extrapolation show that the self-teaching neural networks not only reduce the teaching efforts of the human, but the genetically created examples also contribute robustly to the improvement of generalization performance and the interpretation of the connectionist knowledge.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114858640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170430
I. Y. Poteryaiko
The author proposes a modification of the neural network model of B. Baird (1988,1989) in which the constraint of symmetrical interaction between the modes representing the patterns stored is eliminated. This makes it possible to construct the system with the ordered transitions between the patterns which were the stable attractors in the original model. Although in this case there is no strict evidence that the system does not have the chaotic behavior, a qualitative investigation and extensive numerical simulations show that the dynamics of the system can be described quite simply in terms of effective excitation wandering through the closed loop. Such motion implies the consequent activation of the static or periodic patterns stored in the network. Thus, it is shown that the model can exhibit more complex, but still programmable, behavior than was originally assumed by B. Baird.<>
{"title":"A dynamical network capable of storing sequences of static or periodic patterns","authors":"I. Y. Poteryaiko","doi":"10.1109/IJCNN.1991.170430","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170430","url":null,"abstract":"The author proposes a modification of the neural network model of B. Baird (1988,1989) in which the constraint of symmetrical interaction between the modes representing the patterns stored is eliminated. This makes it possible to construct the system with the ordered transitions between the patterns which were the stable attractors in the original model. Although in this case there is no strict evidence that the system does not have the chaotic behavior, a qualitative investigation and extensive numerical simulations show that the dynamics of the system can be described quite simply in terms of effective excitation wandering through the closed loop. Such motion implies the consequent activation of the static or periodic patterns stored in the network. Thus, it is shown that the model can exhibit more complex, but still programmable, behavior than was originally assumed by B. Baird.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126289415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170506
Y. Chu, C. M. Hsieh
The authors use a neural technique to implement a positive logic Boolean function or truth table. The neural technique is a perceptron training algorithm by which a Boolean function or truth table can be generated. The connected weight value in the neural network represents the sum of product terms of a Boolean function or row vectors of a truth table. A neural technique for generating functional-link cells for successful learning is described. The authors then provide an improved algorithm to describe the successful learning steps to generate the logic function and then present examples to illustrate these learning steps. Finally, a function diagram is specified to illustrate the overall system function.<>
{"title":"A Boolean function generator with learning capability","authors":"Y. Chu, C. M. Hsieh","doi":"10.1109/IJCNN.1991.170506","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170506","url":null,"abstract":"The authors use a neural technique to implement a positive logic Boolean function or truth table. The neural technique is a perceptron training algorithm by which a Boolean function or truth table can be generated. The connected weight value in the neural network represents the sum of product terms of a Boolean function or row vectors of a truth table. A neural technique for generating functional-link cells for successful learning is described. The authors then provide an improved algorithm to describe the successful learning steps to generate the logic function and then present examples to illustrate these learning steps. Finally, a function diagram is specified to illustrate the overall system function.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128006334","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170672
R. Togneri, Y. Attikiouzel
A parallel implementation of the Kohonen algorithm is proposed using partitioning of the network. This allows an exact implementation of the Kohonen algorithm as opposed to partitioning the data. By using a simple routing strategy the parallel Kohonen algorithm was tested on a PC-based transputer network without the need for any special distributed operating system. The execution time was measured for networks of different size and a varying number of transputers. The execution time decreased as the number of transputers increased. However, for comparatively small-size neural networks the communication overhead caused the execution time to increase when more transputers were used. Thus, the proposed parallel implementation of the Kohonen algorithm is not suitable for massively parallel architectures. It is concluded that in excess of 12 transputers can be used with network sizes of the order of 3000 neurons or more but no more than six transputers can be used with network sizes of the order of 120 neurons.<>
{"title":"Parallel implementation of the Kohonen algorithm on transputer","authors":"R. Togneri, Y. Attikiouzel","doi":"10.1109/IJCNN.1991.170672","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170672","url":null,"abstract":"A parallel implementation of the Kohonen algorithm is proposed using partitioning of the network. This allows an exact implementation of the Kohonen algorithm as opposed to partitioning the data. By using a simple routing strategy the parallel Kohonen algorithm was tested on a PC-based transputer network without the need for any special distributed operating system. The execution time was measured for networks of different size and a varying number of transputers. The execution time decreased as the number of transputers increased. However, for comparatively small-size neural networks the communication overhead caused the execution time to increase when more transputers were used. Thus, the proposed parallel implementation of the Kohonen algorithm is not suitable for massively parallel architectures. It is concluded that in excess of 12 transputers can be used with network sizes of the order of 3000 neurons or more but no more than six transputers can be used with network sizes of the order of 120 neurons.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128044650","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170529
A. Visa, K. Valkealahti, O. Simula
A novel method to detect and recognize clouds from remote sensing images is introduced. The detection and recognition of clouds are based on textures. The images are partitioned into homogeneously textured regions, and the interpretation of those textures is based on a texture map. This map is created by means of artificial neural network methodology. The use of neural network methods makes it possible to apply an unsupervised learning paradigm to train the map continuously. The texture map is created by a self-organizing process of feature vectors. This is performed in an unsupervised way. The labeling is achieved by a supervised process.<>
{"title":"Cloud detection based on texture segmentation by neural network methods","authors":"A. Visa, K. Valkealahti, O. Simula","doi":"10.1109/IJCNN.1991.170529","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170529","url":null,"abstract":"A novel method to detect and recognize clouds from remote sensing images is introduced. The detection and recognition of clouds are based on textures. The images are partitioned into homogeneously textured regions, and the interpretation of those textures is based on a texture map. This map is created by means of artificial neural network methodology. The use of neural network methods makes it possible to apply an unsupervised learning paradigm to train the map continuously. The texture map is created by a self-organizing process of feature vectors. This is performed in an unsupervised way. The labeling is achieved by a supervised process.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125705855","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170706
K. Murase, Y. Nakade, Y. Matsunaga, O. Yamakawa
The neural cocktail-party processor (NCPP) is known as a theoretical model of the visual binding by coherent oscillation of neurons, a hypothesis that transitory synchronization of neuronal activities might link fragmentarily represented visual information in the widely spaced areas of the brain to establish coherent images. However, NCPP was made under an assumption that the images to be recognized have no common features. If there are any common features the synchronization among cells is disturbed and the network cannot recognize the images correctly. The authors therefore developed a network, called the neural searchlight processor (NSP), that recognizes images by transitory synchronization allowing common features between images in the input pattern. The mechanism and results of computer simulation of NCPP are described. Then the structure and simulation of NSP are explained by comparison with NCPP.<>
{"title":"A neural searchlight processor that differentiates any images with common features by transitory synchronization","authors":"K. Murase, Y. Nakade, Y. Matsunaga, O. Yamakawa","doi":"10.1109/IJCNN.1991.170706","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170706","url":null,"abstract":"The neural cocktail-party processor (NCPP) is known as a theoretical model of the visual binding by coherent oscillation of neurons, a hypothesis that transitory synchronization of neuronal activities might link fragmentarily represented visual information in the widely spaced areas of the brain to establish coherent images. However, NCPP was made under an assumption that the images to be recognized have no common features. If there are any common features the synchronization among cells is disturbed and the network cannot recognize the images correctly. The authors therefore developed a network, called the neural searchlight processor (NSP), that recognizes images by transitory synchronization allowing common features between images in the input pattern. The mechanism and results of computer simulation of NCPP are described. Then the structure and simulation of NSP are explained by comparison with NCPP.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127937269","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170783
Chung-Mong Lee, D. W. Patterson
A system which combines the power of neural network learning and computing with conventional vision processing methods has been developed. At the heart of the system is a neural network composed of neocognitron and self-created layer components. During the recognition phase, the network computations are augmented by conventional vision algorithms which perform some low- and intermediate-level processing functions. The system is first trained under supervision to recognize several types of nonoccluded objects. It is then used to identify each of the objects appearing in an image even though the objects appear at different locations and are partially occluded or even somewhat deformed. A high degree of accuracy has been achieved with the system.<>
{"title":"Occluded object recognition: an approach which combines neurocomputing and conventional algorithms","authors":"Chung-Mong Lee, D. W. Patterson","doi":"10.1109/IJCNN.1991.170783","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170783","url":null,"abstract":"A system which combines the power of neural network learning and computing with conventional vision processing methods has been developed. At the heart of the system is a neural network composed of neocognitron and self-created layer components. During the recognition phase, the network computations are augmented by conventional vision algorithms which perform some low- and intermediate-level processing functions. The system is first trained under supervision to recognize several types of nonoccluded objects. It is then used to identify each of the objects appearing in an image even though the objects appear at different locations and are partially occluded or even somewhat deformed. A high degree of accuracy has been achieved with the system.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115793690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170652
Y.B. Cho, K.C. Lee, Yoshiyasu Takefuji, N. Funabiki
The circuit of the maximum neural network based on the switched capacitor technique is proposed. The performance of the proposed circuit was derived from SPICE simulation. The bipartite subgraph problem is solved by using the proposed circuit. The SPICE simulation result confirms the function of the network. Because the complexity of the proposed analog circuit is so small, it is possible to fabricate an optimization system in a single chip.<>
{"title":"Analog maximum neural network circuits using the switched capacitor technique","authors":"Y.B. Cho, K.C. Lee, Yoshiyasu Takefuji, N. Funabiki","doi":"10.1109/IJCNN.1991.170652","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170652","url":null,"abstract":"The circuit of the maximum neural network based on the switched capacitor technique is proposed. The performance of the proposed circuit was derived from SPICE simulation. The bipartite subgraph problem is solved by using the proposed circuit. The SPICE simulation result confirms the function of the network. Because the complexity of the proposed analog circuit is so small, it is possible to fabricate an optimization system in a single chip.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132065789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-18DOI: 10.1109/IJCNN.1991.170618
J. Hakala, G. Fahner, R. Eckmiller
An adaptive neural structure for robot control based on homogeneous encoding in a topographical manner is developed. An intermediate representation (IRep) is adaptively generated using a novel learning scheme, CATE. The connection assignment rules of CATE keep the number of IRep-neurons as small as possible, while maintaining the desired mapping accuracy. This adaptive net (CATEnet) was successfully applied to embed the inverse kinematics of a redundant, planar robot arm (four-joint-machine) with only a few presentations of the learning set. The mapping solution incorporated local optimization of a cost function to account for a limited joint range and to avoid singularities.<>
{"title":"Rapid learning of inverse robot kinematics based on connection assignment and topographical encoding (CATE)","authors":"J. Hakala, G. Fahner, R. Eckmiller","doi":"10.1109/IJCNN.1991.170618","DOIUrl":"https://doi.org/10.1109/IJCNN.1991.170618","url":null,"abstract":"An adaptive neural structure for robot control based on homogeneous encoding in a topographical manner is developed. An intermediate representation (IRep) is adaptively generated using a novel learning scheme, CATE. The connection assignment rules of CATE keep the number of IRep-neurons as small as possible, while maintaining the desired mapping accuracy. This adaptive net (CATEnet) was successfully applied to embed the inverse kinematics of a redundant, planar robot arm (four-joint-machine) with only a few presentations of the learning set. The mapping solution incorporated local optimization of a cost function to account for a limited joint range and to avoid singularities.<<ETX>>","PeriodicalId":211135,"journal":{"name":"[Proceedings] 1991 IEEE International Joint Conference on Neural Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1991-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132136312","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：